Our live study has finished!
Visit 3-D Gallery

(See the pictures uploaded by tens of thousands of users.)


Download code, data, and also see all the media releases.



 

Make3D converts your still picture into a 3D model completely automatically---upload, wait for a few seconds, and download!

It takes a two-dimensional image and creates a three-dimensional "fly around" model, giving the viewers access to the scene's depth and a range of points of view. After uploading your image, you can "fly" in the 3-D scene (requires VRML viewer, Flash or Adobe Shockwave), or watch a rendered 3-d movie (flash required). Visit 3-D Gallery or Hall of fame. Also see Google Android version.

How does it work?
It uses powerful machine learning techniques to learn the 3-d structure of a scene as a function of the (single) image features. The project was led by Prof. Ashutosh Saxena and Prof. Andrew Y. Ng, and now is at Cornell University.

We have now used ideas from Make3D for Robotic Scene Understanding.

 

Photograph 3-D Model
Single image 3-d model snapshot
 
 
image
image

Make3D was created by Prof. Ashutosh Saxena, Prof. Andrew Y. Ng, and other team members of the Stanford 3D Reconstruction Group.

Digg it | Reddit it | Slashdot